Various research studies have been conducted on the topic of students’ conceptions of artificial intelligence (AI), but there is insufficient research into how these conceptions can be changed. In computer science education, there is a need to empirically research solutions or activities for conceptions that have already been discovered in order to make them usable in terms of educational reconstruction and integrate them into instruction. However, this is hardly ever done yet. Our work addresses this issue and can be transferred as a prototype study to other areas of conceptions research in computer science education. To this end, a teaching intervention was conducted with 10th and 11th grade secondary school students in Germany in order to measure the change in their agreement to specific conceptions with the help of pre- and post-tests. This study was conducted with N = 76 students who were divided into an experimental and control group. The results of the study show that conceptual change texts are a promising teaching method for expanding students’ conceptions of AI. This is indicated by the results of the post-test, in which students who worked with the conceptual change text were able to demonstrate greater agreement to the conception presented. However, the results of the study needs to be critically discussed in terms of validity and future perspectives.